Add RWF_DONTCACHE as a read operation flag, which means that any data read
wil be removed from the page cache upon completion. Uses the page cache
to synchronize, and simply prunes folios that were instantiated when the
operation completes. While it would be possible to use private pages for
this, using the page cache as synchronization is handy for a variety of
reasons:
1) No special truncate magic is needed
2) Async buffered reads need some place to serialize, using the page
cache is a lot easier than writing extra code for this
3) The pruning cost is pretty reasonable
and the code to support this is much simpler as a result.
You can think of uncached buffered IO as being the much more attractive
cousin of O_DIRECT - it has none of the restrictions of O_DIRECT. Yes, it
will copy the data, but unlike regular buffered IO, it doesn't run into
the unpredictability of the page cache in terms of reclaim. As an
example, on a test box with 32 drives, reading them with buffered IO looks
as follows:
Reading bs 65536, uncached 0
1s: 145945MB/sec
2s: 158067MB/sec
3s: 157007MB/sec
4s: 148622MB/sec
5s: 118824MB/sec
6s: 70494MB/sec
7s: 41754MB/sec
8s: 90811MB/sec
9s: 92204MB/sec
10s: 95178MB/sec
11s: 95488MB/sec
12s: 95552MB/sec
13s: 96275MB/sec
where it's quite easy to see where the page cache filled up, and
performance went from good to erratic, and finally settles at a much
lower rate. Looking at top while this is ongoing, we see:
PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND
7535 root 20 0 267004 0 0 S 3199 0.0 8:40.65 uncached
3326 root 20 0 0 0 0 R 100.0 0.0 0:16.40 kswapd4
3327 root 20 0 0 0 0 R 100.0 0.0 0:17.22 kswapd5
3328 root 20 0 0 0 0 R 100.0 0.0 0:13.29 kswapd6
3332 root 20 0 0 0 0 R 100.0 0.0 0:11.11 kswapd10
3339 root 20 0 0 0 0 R 100.0 0.0 0:16.25 kswapd17
3348 root 20 0 0 0 0 R 100.0 0.0 0:16.40 kswapd26
3343 root 20 0 0 0 0 R 100.0 0.0 0:16.30 kswapd21
3344 root 20 0 0 0 0 R 100.0 0.0 0:11.92 kswapd22
3349 root 20 0 0 0 0 R 100.0 0.0 0:16.28 kswapd27
3352 root 20 0 0 0 0 R 99.7 0.0 0:11.89 kswapd30
3353 root 20 0 0 0 0 R 96.7 0.0 0:16.04 kswapd31
3329 root 20 0 0 0 0 R 96.4 0.0 0:11.41 kswapd7
3345 root 20 0 0 0 0 R 96.4 0.0 0:13.40 kswapd23
3330 root 20 0 0 0 0 S 91.1 0.0 0:08.28 kswapd8
3350 root 20 0 0 0 0 S 86.8 0.0 0:11.13 kswapd28
3325 root 20 0 0 0 0 S 76.3 0.0 0:07.43 kswapd3
3341 root 20 0 0 0 0 S 74.7 0.0 0:08.85 kswapd19
3334 root 20 0 0 0 0 S 71.7 0.0 0:10.04 kswapd12
3351 root 20 0 0 0 0 R 60.5 0.0 0:09.59 kswapd29
3323 root 20 0 0 0 0 R 57.6 0.0 0:11.50 kswapd1
[...]
which is just showing a partial list of the 32 kswapd threads that are
running mostly full tilt, burning ~28 full CPU cores.
If the same test case is run with RWF_DONTCACHE set for the buffered read,
the output looks as follows:
Reading bs 65536, uncached 0
1s: 153144MB/sec
2s: 156760MB/sec
3s: 158110MB/sec
4s: 158009MB/sec
5s: 158043MB/sec
6s: 157638MB/sec
7s: 157999MB/sec
8s: 158024MB/sec
9s: 157764MB/sec
10s: 157477MB/sec
11s: 157417MB/sec
12s: 157455MB/sec
13s: 157233MB/sec
14s: 156692MB/sec
which is just chugging along at ~155GB/sec of read performance. Looking
at top, we see:
PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND
7961 root 20 0 267004 0 0 S 3180 0.0 5:37.95 uncached
8024 axboe 20 0 14292 4096 0 R 1.0 0.0 0:00.13 top
where just the test app is using CPU, no reclaim is taking place outside
of the main thread. Not only is performance 65% better, it's also using
half the CPU to do it.
Link: https://lkml.kernel.org/r/20241220154831.1086649-9-axboe@kernel.dk
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Cc: Brian Foster <bfoster@redhat.com>
Cc: Chris Mason <clm@meta.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
With the aging feedback no longer considering the distribution of folios
in each generation, rework workingset protection to better distribute
folios across MAX_NR_GENS. This is achieved by reusing PG_workingset and
PG_referenced/LRU_REFS_FLAGS in a slightly different way.
For folios accessed multiple times through file descriptors, make
lru_gen_inc_refs() set additional bits of LRU_REFS_WIDTH in folio->flags
after PG_referenced, then PG_workingset after LRU_REFS_WIDTH. After all
its bits are set, i.e., LRU_REFS_FLAGS|BIT(PG_workingset), a folio is
lazily promoted into the second oldest generation in the eviction path.
And when folio_inc_gen() does that, it clears LRU_REFS_FLAGS so that
lru_gen_inc_refs() can start over. For this case, LRU_REFS_MASK is only
valid when PG_referenced is set.
For folios accessed multiple times through page tables, folio_update_gen()
from a page table walk or lru_gen_set_refs() from a rmap walk sets
PG_referenced after the accessed bit is cleared for the first time.
Thereafter, those two paths set PG_workingset and promote folios to the
youngest generation. Like folio_inc_gen(), when folio_update_gen() does
that, it also clears PG_referenced. For this case, LRU_REFS_MASK is not
used.
For both of the cases, after PG_workingset is set on a folio, it remains
until this folio is either reclaimed, or "deactivated" by
lru_gen_clear_refs(). It can be set again if lru_gen_test_recent()
returns true upon a refault.
When adding folios to the LRU lists, lru_gen_folio_seq() distributes
them as follows:
+---------------------------------+---------------------------------+
| Accessed thru page tables | Accessed thru file descriptors |
+---------------------------------+---------------------------------+
| PG_active (set while isolated) | |
+----------------+----------------+----------------+----------------+
| PG_workingset | PG_referenced | PG_workingset | LRU_REFS_FLAGS |
+---------------------------------+---------------------------------+
|<--------- MIN_NR_GENS --------->| |
|<-------------------------- MAX_NR_GENS -------------------------->|
After this patch, some typical client and server workloads showed
improvements under heavy memory pressure. For example, Python TPC-C,
which was used to benchmark a different approach [1] to better detect
refault distances, showed a significant decrease in total refaults:
Before After Change
Time (seconds) 10801 10801 0%
Executed (transactions) 41472 43663 +5%
workingset_nodes 109070 120244 +10%
workingset_refault_anon 5019627 7281831 +45%
workingset_refault_file 1294678786 554855564 -57%
workingset_refault_total 1299698413 562137395 -57%
[1] https://lore.kernel.org/20230920190244.16839-1-ryncsn@gmail.com/
Link: https://lkml.kernel.org/r/20241231043538.4075764-7-yuzhao@google.com
Signed-off-by: Yu Zhao <yuzhao@google.com>
Reported-by: Kairui Song <kasong@tencent.com>
Closes: https://lore.kernel.org/CAOUHufahuWcKf5f1Sg3emnqX+cODuR=2TQo7T4Gr-QYLujn4RA@mail.gmail.com/
Tested-by: Kalesh Singh <kaleshsingh@google.com>
Cc: Barry Song <v-songbaohua@oppo.com>
Cc: Bharata B Rao <bharata@amd.com>
Cc: David Stevens <stevensd@chromium.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Do not shuffle a folio in the deactivation paths if it is already in the
oldest generation. This reduces the LRU lock contention.
Before this patch, the contention is reproducible by FIO, e.g.,
fio -filename=/dev/nvme1n1p2 -direct=0 -thread -size=1024G \
-rwmixwrite=30 --norandommap --randrepeat=0 -ioengine=sync \
-bs=4k -numjobs=400 -runtime=25000 --time_based \
-group_reporting -name=mglru
98.96%--_raw_spin_lock_irqsave
folio_lruvec_lock_irqsave
|
--98.78%--folio_batch_move_lru
|
--98.63%--deactivate_file_folio
mapping_try_invalidate
invalidate_mapping_pages
invalidate_bdev
blkdev_common_ioctl
blkdev_ioctl
After this patch, deactivate_file_folio() bails out early without taking
the LRU lock.
A side effect is that a folio can be left at the head of the oldest
generation, rather than the tail. If reclaim happens at the same time, it
cannot reclaim this folio immediately. Since there is no known
correlation between truncation and reclaim, this side effect is considered
insignificant.
Link: https://lkml.kernel.org/r/20241231043538.4075764-3-yuzhao@google.com
Reported-by: Bharata B Rao <bharata@amd.com>
Closes: https://lore.kernel.org/CAOUHufawNerxqLm7L9Yywp3HJFiYVrYO26ePUb1jH-qxNGWzyA@mail.gmail.com/
Signed-off-by: Yu Zhao <yuzhao@google.com>
Tested-by: Kalesh Singh <kaleshsingh@google.com>
Cc: Barry Song <v-songbaohua@oppo.com>
Cc: David Stevens <stevensd@chromium.org>
Cc: Kairui Song <kasong@tencent.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Pull MM updates from Andrew Morton:
- The series "zram: optimal post-processing target selection" from
Sergey Senozhatsky improves zram's post-processing selection
algorithm. This leads to improved memory savings.
- Wei Yang has gone to town on the mapletree code, contributing several
series which clean up the implementation:
- "refine mas_mab_cp()"
- "Reduce the space to be cleared for maple_big_node"
- "maple_tree: simplify mas_push_node()"
- "Following cleanup after introduce mas_wr_store_type()"
- "refine storing null"
- The series "selftests/mm: hugetlb_fault_after_madv improvements" from
David Hildenbrand fixes this selftest for s390.
- The series "introduce pte_offset_map_{ro|rw}_nolock()" from Qi Zheng
implements some rationaizations and cleanups in the page mapping
code.
- The series "mm: optimize shadow entries removal" from Shakeel Butt
optimizes the file truncation code by speeding up the handling of
shadow entries.
- The series "Remove PageKsm()" from Matthew Wilcox completes the
migration of this flag over to being a folio-based flag.
- The series "Unify hugetlb into arch_get_unmapped_area functions" from
Oscar Salvador implements a bunch of consolidations and cleanups in
the hugetlb code.
- The series "Do not shatter hugezeropage on wp-fault" from Dev Jain
takes away the wp-fault time practice of turning a huge zero page
into small pages. Instead we replace the whole thing with a THP. More
consistent cleaner and potentiall saves a large number of pagefaults.
- The series "percpu: Add a test case and fix for clang" from Andy
Shevchenko enhances and fixes the kernel's built in percpu test code.
- The series "mm/mremap: Remove extra vma tree walk" from Liam Howlett
optimizes mremap() by avoiding doing things which we didn't need to
do.
- The series "Improve the tmpfs large folio read performance" from
Baolin Wang teaches tmpfs to copy data into userspace at the folio
size rather than as individual pages. A 20% speedup was observed.
- The series "mm/damon/vaddr: Fix issue in
damon_va_evenly_split_region()" fro Zheng Yejian fixes DAMON
splitting.
- The series "memcg-v1: fully deprecate charge moving" from Shakeel
Butt removes the long-deprecated memcgv2 charge moving feature.
- The series "fix error handling in mmap_region() and refactor" from
Lorenzo Stoakes cleanup up some of the mmap() error handling and
addresses some potential performance issues.
- The series "x86/module: use large ROX pages for text allocations"
from Mike Rapoport teaches x86 to use large pages for
read-only-execute module text.
- The series "page allocation tag compression" from Suren Baghdasaryan
is followon maintenance work for the new page allocation profiling
feature.
- The series "page->index removals in mm" from Matthew Wilcox remove
most references to page->index in mm/. A slow march towards shrinking
struct page.
- The series "damon/{self,kunit}tests: minor fixups for DAMON debugfs
interface tests" from Andrew Paniakin performs maintenance work for
DAMON's self testing code.
- The series "mm: zswap swap-out of large folios" from Kanchana Sridhar
improves zswap's batching of compression and decompression. It is a
step along the way towards using Intel IAA hardware acceleration for
this zswap operation.
- The series "kasan: migrate the last module test to kunit" from
Sabyrzhan Tasbolatov completes the migration of the KASAN built-in
tests over to the KUnit framework.
- The series "implement lightweight guard pages" from Lorenzo Stoakes
permits userapace to place fault-generating guard pages within a
single VMA, rather than requiring that multiple VMAs be created for
this. Improved efficiencies for userspace memory allocators are
expected.
- The series "memcg: tracepoint for flushing stats" from JP Kobryn uses
tracepoints to provide increased visibility into memcg stats flushing
activity.
- The series "zram: IDLE flag handling fixes" from Sergey Senozhatsky
fixes a zram buglet which potentially affected performance.
- The series "mm: add more kernel parameters to control mTHP" from
Maíra Canal enhances our ability to control/configuremultisize THP
from the kernel boot command line.
- The series "kasan: few improvements on kunit tests" from Sabyrzhan
Tasbolatov has a couple of fixups for the KASAN KUnit tests.
- The series "mm/list_lru: Split list_lru lock into per-cgroup scope"
from Kairui Song optimizes list_lru memory utilization when lockdep
is enabled.
* tag 'mm-stable-2024-11-18-19-27' of git://git.kernel.org/pub/scm/linux/kernel/git/akpm/mm: (215 commits)
cma: enforce non-zero pageblock_order during cma_init_reserved_mem()
mm/kfence: add a new kunit test test_use_after_free_read_nofault()
zram: fix NULL pointer in comp_algorithm_show()
memcg/hugetlb: add hugeTLB counters to memcg
vmstat: call fold_vm_zone_numa_events() before show per zone NUMA event
mm: mmap_lock: check trace_mmap_lock_$type_enabled() instead of regcount
zram: ZRAM_DEF_COMP should depend on ZRAM
MAINTAINERS/MEMORY MANAGEMENT: add document files for mm
Docs/mm/damon: recommend academic papers to read and/or cite
mm: define general function pXd_init()
kmemleak: iommu/iova: fix transient kmemleak false positive
mm/list_lru: simplify the list_lru walk callback function
mm/list_lru: split the lock to per-cgroup scope
mm/list_lru: simplify reparenting and initial allocation
mm/list_lru: code clean up for reparenting
mm/list_lru: don't export list_lru_add
mm/list_lru: don't pass unnecessary key parameters
kasan: add kunit tests for kmalloc_track_caller, kmalloc_node_track_caller
kasan: change kasan_atomics kunit test as KUNIT_CASE_SLOW
kasan: use EXPORT_SYMBOL_IF_KUNIT to export symbols
...
Recent changes are putting more pressure on THP deferred split queues:
under load revealing long-standing races, causing list_del corruptions,
"Bad page state"s and worse (I keep BUGs in both of those, so usually
don't get to see how badly they end up without). The relevant recent
changes being 6.8's mTHP, 6.10's mTHP swapout, and 6.12's mTHP swapin,
improved swap allocation, and underused THP splitting.
Before fixing locking: rename misleading folio_undo_large_rmappable(),
which does not undo large_rmappable, to folio_unqueue_deferred_split(),
which is what it does. But that and its out-of-line __callee are mm
internals of very limited usability: add comment and WARN_ON_ONCEs to
check usage; and return a bool to say if a deferred split was unqueued,
which can then be used in WARN_ON_ONCEs around safety checks (sparing
callers the arcane conditionals in __folio_unqueue_deferred_split()).
Just omit the folio_unqueue_deferred_split() from free_unref_folios(), all
of whose callers now call it beforehand (and if any forget then bad_page()
will tell) - except for its caller put_pages_list(), which itself no
longer has any callers (and will be deleted separately).
Swapout: mem_cgroup_swapout() has been resetting folio->memcg_data 0
without checking and unqueueing a THP folio from deferred split list;
which is unfortunate, since the split_queue_lock depends on the memcg
(when memcg is enabled); so swapout has been unqueueing such THPs later,
when freeing the folio, using the pgdat's lock instead: potentially
corrupting the memcg's list. __remove_mapping() has frozen refcount to 0
here, so no problem with calling folio_unqueue_deferred_split() before
resetting memcg_data.
That goes back to 5.4 commit 87eaceb3fa ("mm: thp: make deferred split
shrinker memcg aware"): which included a check on swapcache before adding
to deferred queue, but no check on deferred queue before adding THP to
swapcache. That worked fine with the usual sequence of events in reclaim
(though there were a couple of rare ways in which a THP on deferred queue
could have been swapped out), but 6.12 commit dafff3f4c8 ("mm: split
underused THPs") avoids splitting underused THPs in reclaim, which makes
swapcache THPs on deferred queue commonplace.
Keep the check on swapcache before adding to deferred queue? Yes: it is
no longer essential, but preserves the existing behaviour, and is likely
to be a worthwhile optimization (vmstat showed much more traffic on the
queue under swapping load if the check was removed); update its comment.
Memcg-v1 move (deprecated): mem_cgroup_move_account() has been changing
folio->memcg_data without checking and unqueueing a THP folio from the
deferred list, sometimes corrupting "from" memcg's list, like swapout.
Refcount is non-zero here, so folio_unqueue_deferred_split() can only be
used in a WARN_ON_ONCE to validate the fix, which must be done earlier:
mem_cgroup_move_charge_pte_range() first try to split the THP (splitting
of course unqueues), or skip it if that fails. Not ideal, but moving
charge has been requested, and khugepaged should repair the THP later:
nobody wants new custom unqueueing code just for this deprecated case.
The 87eaceb3fa commit did have the code to move from one deferred list
to another (but was not conscious of its unsafety while refcount non-0);
but that was removed by 5.6 commit fac0516b55 ("mm: thp: don't need care
deferred split queue in memcg charge move path"), which argued that the
existence of a PMD mapping guarantees that the THP cannot be on a deferred
list. As above, false in rare cases, and now commonly false.
Backport to 6.11 should be straightforward. Earlier backports must take
care that other _deferred_list fixes and dependencies are included. There
is not a strong case for backports, but they can fix cornercases.
Link: https://lkml.kernel.org/r/8dc111ae-f6db-2da7-b25c-7a20b1effe3b@google.com
Fixes: 87eaceb3fa ("mm: thp: make deferred split shrinker memcg aware")
Fixes: dafff3f4c8 ("mm: split underused THPs")
Signed-off-by: Hugh Dickins <hughd@google.com>
Acked-by: David Hildenbrand <david@redhat.com>
Reviewed-by: Yang Shi <shy828301@gmail.com>
Cc: Baolin Wang <baolin.wang@linux.alibaba.com>
Cc: Barry Song <baohua@kernel.org>
Cc: Chris Li <chrisl@kernel.org>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Kefeng Wang <wangkefeng.wang@huawei.com>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Nhat Pham <nphamcs@gmail.com>
Cc: Ryan Roberts <ryan.roberts@arm.com>
Cc: Shakeel Butt <shakeel.butt@linux.dev>
Cc: Usama Arif <usamaarif642@gmail.com>
Cc: Wei Yang <richard.weiyang@gmail.com>
Cc: Zi Yan <ziy@nvidia.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Fold lru_rotate into cpu_fbatches, and rename the folio_batch and the lock
protecting it to lru_move_tail and lock_irq respectively so that all the
boilerplate can be removed at the end of this series.
Also remove data_race() around folio_batch_count(), which is out of place:
all folio_batch_count() calls on remote cpu_fbatches are subject to
data_race(), and therefore data_race() should be inside
folio_batch_count().
Link: https://lkml.kernel.org/r/20240711021317.596178-4-yuzhao@google.com
Signed-off-by: Yu Zhao <yuzhao@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Patch series "mm/swap: remove boilerplate".
This patch (of 5):
Use folio_activate() as an example:
Before this series
------------------
if (!folio_test_active(folio) && !folio_test_unevictable(folio)) {
struct folio_batch *fbatch;
folio_get(folio);
if (!folio_test_clear_lru(folio)) {
folio_put(folio);
return;
}
local_lock(&cpu_fbatches.lock);
fbatch = this_cpu_ptr(&cpu_fbatches.activate);
folio_batch_add_and_move(fbatch, folio, folio_activate_fn);
local_unlock(&cpu_fbatches.lock);
}
}
After this series
-----------------
void folio_activate(struct folio *folio)
{
if (folio_test_active(folio) || folio_test_unevictable(folio))
return;
folio_batch_add_and_move(folio, lru_activate, true);
}
And this is applied to all 6 folio_batch handlers in mm/swap.c.
bloat-o-meter
-------------
add/remove: 12/13 grow/shrink: 3/2 up/down: 4653/-4721 (-68)
...
Total: Before=28083019, After=28082951, chg -0.00%
This patch (of 5):
Reduce indentation level by returning directly when there is no cleanup
needed, i.e.,
if (condition) { | if (condition) {
do_this(); | do_this();
return; | return;
} else { | }
do_that(); |
} | do_that();
and
if (condition) { | if (!condition)
do_this(); | return;
do_that(); |
} | do_this();
return; | do_that();
Presumably the old style became repetitive as the result of copy and
paste.
Link: https://lkml.kernel.org/r/20240711021317.596178-1-yuzhao@google.com
Link: https://lkml.kernel.org/r/20240711021317.596178-2-yuzhao@google.com
Signed-off-by: Yu Zhao <yuzhao@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
If a large number of CMA memory are configured in system (for example,
the CMA memory accounts for 50% of the system memory), starting a
virtual virtual machine with device passthrough, it will call
pin_user_pages_remote(..., FOLL_LONGTERM, ...) to pin memory. Normally
if a page is present and in CMA area, pin_user_pages_remote() will
migrate the page from CMA area to non-CMA area because of FOLL_LONGTERM
flag. But the current code will cause the migration failure due to
unexpected page refcounts, and eventually cause the virtual machine
fail to start.
If a page is added in LRU batch, its refcount increases one, remove the
page from LRU batch decreases one. Page migration requires the page is
not referenced by others except page mapping. Before migrating a page,
we should try to drain the page from LRU batch in case the page is in
it, however, folio_test_lru() is not sufficient to tell whether the
page is in LRU batch or not, if the page is in LRU batch, the migration
will fail.
To solve the problem above, we modify the logic of adding to LRU batch.
Before adding a page to LRU batch, we clear the LRU flag of the page
so that we can check whether the page is in LRU batch by
folio_test_lru(page). It's quite valuable, because likely we don't
want to blindly drain the LRU batch simply because there is some
unexpected reference on a page, as described above.
This change makes the LRU flag of a page invisible for longer, which
may impact some programs. For example, as long as a page is on a LRU
batch, we cannot isolate it, and we cannot check if it's an LRU page.
Further, a page can now only be on exactly one LRU batch. This doesn't
seem to matter much, because a new page is allocated from buddy and
added to the lru batch, or be isolated, it's LRU flag may also be
invisible for a long time.
Link: https://lkml.kernel.org/r/1720075944-27201-1-git-send-email-yangge1116@126.com
Link: https://lkml.kernel.org/r/1720008153-16035-1-git-send-email-yangge1116@126.com
Fixes: 9a4e9f3b2d ("mm: update get_user_pages_longterm to migrate pages allocated from CMA region")
Signed-off-by: yangge <yangge1116@126.com>
Cc: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com>
Cc: Baolin Wang <baolin.wang@linux.alibaba.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Barry Song <21cnbao@gmail.com>
Cc: Hugh Dickins <hughd@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Patch series "Rearrange batched folio freeing", v3.
Other than the obvious "remove calls to compound_head" changes, the
fundamental belief here is that iterating a linked list is much slower
than iterating an array (5-15x slower in my testing). There's also an
associated belief that since we iterate the batch of folios three times,
we do better when the array is small (ie 15 entries) than we do with a
batch that is hundreds of entries long, which only gives us the
opportunity for the first pages to fall out of cache by the time we get to
the end.
It is possible we should increase the size of folio_batch. Hopefully the
bots let us know if this introduces any performance regressions.
This patch (of 3):
By making release_pages() call folios_put(), we can get rid of the calls
to compound_head() for the callers that already know they have folios. We
can also get rid of the lock_batch tracking as we know the size of the
batch is limited by folio_batch. This does reduce the maximum number of
pages for which the lruvec lock is held, from SWAP_CLUSTER_MAX (32) to
PAGEVEC_SIZE (15). I do not expect this to make a significant difference,
but if it does, we can increase PAGEVEC_SIZE to 31.
Link: https://lkml.kernel.org/r/20240227174254.710559-1-willy@infradead.org
Link: https://lkml.kernel.org/r/20240227174254.710559-2-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: David Hildenbrand <david@redhat.com>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Ryan Roberts <ryan.roberts@arm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Add __tlb_remove_folio_pages(), which will remove multiple consecutive
pages that belong to the same large folio, instead of only a single page.
We'll be using this function when optimizing unmapping/zapping of large
folios that are mapped by PTEs.
We're using the remaining spare bit in an encoded_page to indicate that
the next enoced page in an array contains actually shifted "nr_pages".
Teach swap/freeing code about putting multiple folio references, and
delayed rmap handling to remove page ranges of a folio.
This extension allows for still gathering almost as many small folios as
we used to (-1, because we have to prepare for a possibly bigger next
entry), but still allows for gathering consecutive pages that belong to
the same large folio.
Note that we don't pass the folio pointer, because it is not required for
now. Further, we don't support page_size != PAGE_SIZE, it won't be
required for simple PTE batching.
We have to provide a separate s390 implementation, but it's fairly
straight forward.
Another, more invasive and likely more expensive, approach would be to use
folio+range or a PFN range instead of page+nr_pages. But, we should do
that consistently for the whole mmu_gather. For now, let's keep it simple
and add "nr_pages" only.
Note that it is now possible to gather significantly more pages: In the
past, we were able to gather ~10000 pages, now we can also gather ~5000
folio fragments that span multiple pages. A folio fragment on x86-64 can
span up to 512 pages (2 MiB THP) and on arm64 with 64k in theory 8192
pages (512 MiB THP). Gathering more memory is not considered something we
should worry about, especially because these are already corner cases.
While we can gather more total memory, we won't free more folio fragments.
As long as page freeing time primarily only depends on the number of
involved folios, there is no effective change for !preempt configurations.
However, we'll adjust tlb_batch_pages_flush() separately to handle corner
cases where page freeing time grows proportionally with the actual memory
size.
Link: https://lkml.kernel.org/r/20240214204435.167852-9-david@redhat.com
Signed-off-by: David Hildenbrand <david@redhat.com>
Reviewed-by: Ryan Roberts <ryan.roberts@arm.com>
Cc: Alexander Gordeev <agordeev@linux.ibm.com>
Cc: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Christian Borntraeger <borntraeger@linux.ibm.com>
Cc: Christophe Leroy <christophe.leroy@csgroup.eu>
Cc: Heiko Carstens <hca@linux.ibm.com>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Michal Hocko <mhocko@suse.com>
Cc: "Naveen N. Rao" <naveen.n.rao@linux.ibm.com>
Cc: Nicholas Piggin <npiggin@gmail.com>
Cc: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Sven Schnelle <svens@linux.ibm.com>
Cc: Vasily Gorbik <gor@linux.ibm.com>
Cc: Will Deacon <will@kernel.org>
Cc: Yin Fengwei <fengwei.yin@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
The ->percpu_pvec_drained was originally introduced by commit d9ed0d08b6
("mm: only drain per-cpu pagevecs once per pagevec usage") to drain
per-cpu pagevecs only once per pagevec usage. But after converting the
swap code to be more folio-based, the commit c2bc16817a ("mm/swap: add
folio_batch_move_lru()") breaks this logic, which would cause
->percpu_pvec_drained to be reset to false, that means per-cpu pagevecs
will be drained multiple times per pagevec usage.
In theory, there should be no functional changes when converting code to
be more folio-based. We should call folio_batch_reinit() in
folio_batch_move_lru() instead of folio_batch_init(). And to verify that
we still need ->percpu_pvec_drained, I ran mmtests/sparsetruncate-tiny and
got the following data:
baseline with
baseline/ patch/
Min Time 326.00 ( 0.00%) 328.00 ( -0.61%)
1st-qrtle Time 334.00 ( 0.00%) 336.00 ( -0.60%)
2nd-qrtle Time 338.00 ( 0.00%) 341.00 ( -0.89%)
3rd-qrtle Time 343.00 ( 0.00%) 347.00 ( -1.17%)
Max-1 Time 326.00 ( 0.00%) 328.00 ( -0.61%)
Max-5 Time 327.00 ( 0.00%) 330.00 ( -0.92%)
Max-10 Time 328.00 ( 0.00%) 331.00 ( -0.91%)
Max-90 Time 350.00 ( 0.00%) 357.00 ( -2.00%)
Max-95 Time 395.00 ( 0.00%) 390.00 ( 1.27%)
Max-99 Time 508.00 ( 0.00%) 434.00 ( 14.57%)
Max Time 547.00 ( 0.00%) 476.00 ( 12.98%)
Amean Time 344.61 ( 0.00%) 345.56 * -0.28%*
Stddev Time 30.34 ( 0.00%) 19.51 ( 35.69%)
CoeffVar Time 8.81 ( 0.00%) 5.65 ( 35.87%)
BAmean-99 Time 342.38 ( 0.00%) 344.27 ( -0.55%)
BAmean-95 Time 338.58 ( 0.00%) 341.87 ( -0.97%)
BAmean-90 Time 336.89 ( 0.00%) 340.26 ( -1.00%)
BAmean-75 Time 335.18 ( 0.00%) 338.40 ( -0.96%)
BAmean-50 Time 332.54 ( 0.00%) 335.42 ( -0.87%)
BAmean-25 Time 329.30 ( 0.00%) 332.00 ( -0.82%)
From the above it can be seen that we get similar data to when
->percpu_pvec_drained was introduced, so we still need it. Let's call
folio_batch_reinit() in folio_batch_move_lru() to restore the original
logic.
Link: https://lkml.kernel.org/r/20230405161854.6931-1-zhengqi.arch@bytedance.com
Fixes: c2bc16817a ("mm/swap: add folio_batch_move_lru()")
Signed-off-by: Qi Zheng <zhengqi.arch@bytedance.com>
Reviewed-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Acked-by: Mel Gorman <mgorman@suse.de>
Cc: Lorenzo Stoakes <lstoakes@gmail.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Pull MM updates from Andrew Morton:
- Daniel Verkamp has contributed a memfd series ("mm/memfd: add
F_SEAL_EXEC") which permits the setting of the memfd execute bit at
memfd creation time, with the option of sealing the state of the X
bit.
- Peter Xu adds a patch series ("mm/hugetlb: Make huge_pte_offset()
thread-safe for pmd unshare") which addresses a rare race condition
related to PMD unsharing.
- Several folioification patch serieses from Matthew Wilcox, Vishal
Moola, Sidhartha Kumar and Lorenzo Stoakes
- Johannes Weiner has a series ("mm: push down lock_page_memcg()")
which does perform some memcg maintenance and cleanup work.
- SeongJae Park has added DAMOS filtering to DAMON, with the series
"mm/damon/core: implement damos filter".
These filters provide users with finer-grained control over DAMOS's
actions. SeongJae has also done some DAMON cleanup work.
- Kairui Song adds a series ("Clean up and fixes for swap").
- Vernon Yang contributed the series "Clean up and refinement for maple
tree".
- Yu Zhao has contributed the "mm: multi-gen LRU: memcg LRU" series. It
adds to MGLRU an LRU of memcgs, to improve the scalability of global
reclaim.
- David Hildenbrand has added some userfaultfd cleanup work in the
series "mm: uffd-wp + change_protection() cleanups".
- Christoph Hellwig has removed the generic_writepages() library
function in the series "remove generic_writepages".
- Baolin Wang has performed some maintenance on the compaction code in
his series "Some small improvements for compaction".
- Sidhartha Kumar is doing some maintenance work on struct page in his
series "Get rid of tail page fields".
- David Hildenbrand contributed some cleanup, bugfixing and
generalization of pte management and of pte debugging in his series
"mm: support __HAVE_ARCH_PTE_SWP_EXCLUSIVE on all architectures with
swap PTEs".
- Mel Gorman and Neil Brown have removed the __GFP_ATOMIC allocation
flag in the series "Discard __GFP_ATOMIC".
- Sergey Senozhatsky has improved zsmalloc's memory utilization with
his series "zsmalloc: make zspage chain size configurable".
- Joey Gouly has added prctl() support for prohibiting the creation of
writeable+executable mappings.
The previous BPF-based approach had shortcomings. See "mm: In-kernel
support for memory-deny-write-execute (MDWE)".
- Waiman Long did some kmemleak cleanup and bugfixing in the series
"mm/kmemleak: Simplify kmemleak_cond_resched() & fix UAF".
- T.J. Alumbaugh has contributed some MGLRU cleanup work in his series
"mm: multi-gen LRU: improve".
- Jiaqi Yan has provided some enhancements to our memory error
statistics reporting, mainly by presenting the statistics on a
per-node basis. See the series "Introduce per NUMA node memory error
statistics".
- Mel Gorman has a second and hopefully final shot at fixing a CPU-hog
regression in compaction via his series "Fix excessive CPU usage
during compaction".
- Christoph Hellwig does some vmalloc maintenance work in the series
"cleanup vfree and vunmap".
- Christoph Hellwig has removed block_device_operations.rw_page() in
ths series "remove ->rw_page".
- We get some maple_tree improvements and cleanups in Liam Howlett's
series "VMA tree type safety and remove __vma_adjust()".
- Suren Baghdasaryan has done some work on the maintainability of our
vm_flags handling in the series "introduce vm_flags modifier
functions".
- Some pagemap cleanup and generalization work in Mike Rapoport's
series "mm, arch: add generic implementation of pfn_valid() for
FLATMEM" and "fixups for generic implementation of pfn_valid()"
- Baoquan He has done some work to make /proc/vmallocinfo and
/proc/kcore better represent the real state of things in his series
"mm/vmalloc.c: allow vread() to read out vm_map_ram areas".
- Jason Gunthorpe rationalized the GUP system's interface to the rest
of the kernel in the series "Simplify the external interface for
GUP".
- SeongJae Park wishes to migrate people from DAMON's debugfs interface
over to its sysfs interface. To support this, we'll temporarily be
printing warnings when people use the debugfs interface. See the
series "mm/damon: deprecate DAMON debugfs interface".
- Andrey Konovalov provided the accurately named "lib/stackdepot: fixes
and clean-ups" series.
- Huang Ying has provided a dramatic reduction in migration's TLB flush
IPI rates with the series "migrate_pages(): batch TLB flushing".
- Arnd Bergmann has some objtool fixups in "objtool warning fixes".
* tag 'mm-stable-2023-02-20-13-37' of git://git.kernel.org/pub/scm/linux/kernel/git/akpm/mm: (505 commits)
include/linux/migrate.h: remove unneeded externs
mm/memory_hotplug: cleanup return value handing in do_migrate_range()
mm/uffd: fix comment in handling pte markers
mm: change to return bool for isolate_movable_page()
mm: hugetlb: change to return bool for isolate_hugetlb()
mm: change to return bool for isolate_lru_page()
mm: change to return bool for folio_isolate_lru()
objtool: add UACCESS exceptions for __tsan_volatile_read/write
kmsan: disable ftrace in kmsan core code
kasan: mark addr_has_metadata __always_inline
mm: memcontrol: rename memcg_kmem_enabled()
sh: initialize max_mapnr
m68k/nommu: add missing definition of ARCH_PFN_OFFSET
mm: percpu: fix incorrect size in pcpu_obj_full_size()
maple_tree: reduce stack usage with gcc-9 and earlier
mm: page_alloc: call panic() when memoryless node allocation fails
mm: multi-gen LRU: avoid futile retries
migrate_pages: move THP/hugetlb migration support check to simplify code
migrate_pages: batch flushing TLB
migrate_pages: share more code between _unmap and _move
...
Deactivate_page() has already been converted to use folios, this change
converts it to take in a folio argument instead of calling page_folio().
It also renames the function folio_deactivate() to be more consistent with
other folio functions.
[akpm@linux-foundation.org: fix left-over comments, per Yu Zhao]
Link: https://lkml.kernel.org/r/20221221180848.20774-5-vishal.moola@gmail.com
Signed-off-by: Vishal Moola (Oracle) <vishal.moola@gmail.com>
Reviewed-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: SeongJae Park <sj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
release_pages() already could take either an array of page pointers, or an
array of folio pointers. Expand it to also accept an array of encoded
page pointers, which is what both the existing mlock() use and the
upcoming mmu_gather use of encoded page pointers wants.
Note that release_pages() won't actually use, or react to, any extra
encoded bits. Instead, this is very much a case of "I have walked the
array of encoded pages and done everything the extra bits tell me to do,
now release it all".
Also, while the "either page or folio pointers" dual use was handled with
a cast of the pointer in "release_folios()", this takes a slightly
different approach and uses the "transparent union" attribute to describe
the set of arguments to the function:
https://gcc.gnu.org/onlinedocs/gcc/Common-Type-Attributes.html
which has been supported by gcc forever, but the kernel hasn't used
before.
That allows us to avoid using various wrappers with casts, and just use
the same function regardless of use.
Link: https://lkml.kernel.org/r/20221109203051.1835763-2-torvalds@linux-foundation.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Hugh Dickins <hughd@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Currenty there is no upper limit for /proc/sys/vm/page-cluster, and it's a
bit shift value, so it could result in overflow of the 32-bit integer.
Add a reasonable upper limit for it, read-in at most 2**31 pages, which is
a large enough value for readahead.
Link: https://lkml.kernel.org/r/20221023162533.81561-1-ryncsn@gmail.com
Signed-off-by: Kairui Song <kasong@tencent.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
We noticed a 2% webserver throughput regression after upgrading from 5.6.
This could be tracked down to a shift in the anon/file reclaim balance
(confirmed with swappiness) that resulted in worse reclaim efficiency and
thus more kswapd activity for the same outcome.
The change that exposed the problem is aae466b005 ("mm/swap: implement
workingset detection for anonymous LRU"). By qualifying swapins based on
their refault distance, it lowered the cost of anon reclaim in this
workload, in turn causing (much) more anon scanning than before. Scanning
the anon list is more expensive due to the higher ratio of mmapped pages
that may rotate during reclaim, and so the result was an increase in %sys
time.
Right now, rotations aren't considered a cost when balancing scan pressure
between LRUs. We can end up with very few file refaults putting all the
scan pressure on hot anon pages that are rotated en masse, don't get
reclaimed, and never push back on the file LRU again. We still only
reclaim file cache in that case, but we burn a lot CPU rotating anon
pages. It's "fair" from an LRU age POV, but doesn't reflect the real cost
it imposes on the system.
Consider rotations as a secondary factor in balancing the LRUs. This
doesn't attempt to make a precise comparison between IO cost and CPU cost,
it just says: if reloads are about comparable between the lists, or
rotations are overwhelmingly different, adjust for CPU work.
This fixed the regression on our webservers. It has since been deployed
to the entire Meta fleet and hasn't caused any problems.
Link: https://lkml.kernel.org/r/20221013193113.726425-1-hannes@cmpxchg.org
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Rik van Riel <riel@surriel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Searching the rmap for PTEs mapping each page on an LRU list (to test and
clear the accessed bit) can be expensive because pages from different VMAs
(PA space) are not cache friendly to the rmap (VA space). For workloads
mostly using mapped pages, searching the rmap can incur the highest CPU
cost in the reclaim path.
This patch exploits spatial locality to reduce the trips into the rmap.
When shrink_page_list() walks the rmap and finds a young PTE, a new
function lru_gen_look_around() scans at most BITS_PER_LONG-1 adjacent
PTEs. On finding another young PTE, it clears the accessed bit and
updates the gen counter of the page mapped by this PTE to
(max_seq%MAX_NR_GENS)+1.
Server benchmark results:
Single workload:
fio (buffered I/O): no change
Single workload:
memcached (anon): +[3, 5]%
Ops/sec KB/sec
patch1-6: 1106168.46 43025.04
patch1-7: 1147696.57 44640.29
Configurations:
no change
Client benchmark results:
kswapd profiles:
patch1-6
39.03% lzo1x_1_do_compress (real work)
18.47% page_vma_mapped_walk (overhead)
6.74% _raw_spin_unlock_irq
3.97% do_raw_spin_lock
2.49% ptep_clear_flush
2.48% anon_vma_interval_tree_iter_first
1.92% folio_referenced_one
1.88% __zram_bvec_write
1.48% memmove
1.31% vma_interval_tree_iter_next
patch1-7
48.16% lzo1x_1_do_compress (real work)
8.20% page_vma_mapped_walk (overhead)
7.06% _raw_spin_unlock_irq
2.92% ptep_clear_flush
2.53% __zram_bvec_write
2.11% do_raw_spin_lock
2.02% memmove
1.93% lru_gen_look_around
1.56% free_unref_page_list
1.40% memset
Configurations:
no change
Link: https://lkml.kernel.org/r/20220918080010.2920238-8-yuzhao@google.com
Signed-off-by: Yu Zhao <yuzhao@google.com>
Acked-by: Barry Song <baohua@kernel.org>
Acked-by: Brian Geffon <bgeffon@google.com>
Acked-by: Jan Alexander Steffens (heftig) <heftig@archlinux.org>
Acked-by: Oleksandr Natalenko <oleksandr@natalenko.name>
Acked-by: Steven Barrett <steven@liquorix.net>
Acked-by: Suleiman Souhlal <suleiman@google.com>
Tested-by: Daniel Byrne <djbyrne@mtu.edu>
Tested-by: Donald Carr <d@chaos-reins.com>
Tested-by: Holger Hoffstätte <holger@applied-asynchrony.com>
Tested-by: Konstantin Kharlamov <Hi-Angel@yandex.ru>
Tested-by: Shuang Zhai <szhai2@cs.rochester.edu>
Tested-by: Sofia Trinh <sofia.trinh@edi.works>
Tested-by: Vaibhav Jain <vaibhav@linux.ibm.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Hillf Danton <hdanton@sina.com>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Michael Larabel <Michael@MichaelLarabel.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Mike Rapoport <rppt@kernel.org>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Qi Zheng <zhengqi.arch@bytedance.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Will Deacon <will@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
To avoid confusion, the terms "promotion" and "demotion" will be applied
to the multi-gen LRU, as a new convention; the terms "activation" and
"deactivation" will be applied to the active/inactive LRU, as usual.
The aging produces young generations. Given an lruvec, it increments
max_seq when max_seq-min_seq+1 approaches MIN_NR_GENS. The aging promotes
hot pages to the youngest generation when it finds them accessed through
page tables; the demotion of cold pages happens consequently when it
increments max_seq. Promotion in the aging path does not involve any LRU
list operations, only the updates of the gen counter and
lrugen->nr_pages[]; demotion, unless as the result of the increment of
max_seq, requires LRU list operations, e.g., lru_deactivate_fn(). The
aging has the complexity O(nr_hot_pages), since it is only interested in
hot pages.
The eviction consumes old generations. Given an lruvec, it increments
min_seq when lrugen->lists[] indexed by min_seq%MAX_NR_GENS becomes empty.
A feedback loop modeled after the PID controller monitors refaults over
anon and file types and decides which type to evict when both types are
available from the same generation.
The protection of pages accessed multiple times through file descriptors
takes place in the eviction path. Each generation is divided into
multiple tiers. A page accessed N times through file descriptors is in
tier order_base_2(N). Tiers do not have dedicated lrugen->lists[], only
bits in folio->flags. The aforementioned feedback loop also monitors
refaults over all tiers and decides when to protect pages in which tiers
(N>1), using the first tier (N=0,1) as a baseline. The first tier
contains single-use unmapped clean pages, which are most likely the best
choices. In contrast to promotion in the aging path, the protection of a
page in the eviction path is achieved by moving this page to the next
generation, i.e., min_seq+1, if the feedback loop decides so. This
approach has the following advantages:
1. It removes the cost of activation in the buffered access path by
inferring whether pages accessed multiple times through file
descriptors are statistically hot and thus worth protecting in the
eviction path.
2. It takes pages accessed through page tables into account and avoids
overprotecting pages accessed multiple times through file
descriptors. (Pages accessed through page tables are in the first
tier, since N=0.)
3. More tiers provide better protection for pages accessed more than
twice through file descriptors, when under heavy buffered I/O
workloads.
Server benchmark results:
Single workload:
fio (buffered I/O): +[30, 32]%
IOPS BW
5.19-rc1: 2673k 10.2GiB/s
patch1-6: 3491k 13.3GiB/s
Single workload:
memcached (anon): -[4, 6]%
Ops/sec KB/sec
5.19-rc1: 1161501.04 45177.25
patch1-6: 1106168.46 43025.04
Configurations:
CPU: two Xeon 6154
Mem: total 256G
Node 1 was only used as a ram disk to reduce the variance in the
results.
patch drivers/block/brd.c <<EOF
99,100c99,100
< gfp_flags = GFP_NOIO | __GFP_ZERO | __GFP_HIGHMEM;
< page = alloc_page(gfp_flags);
---
> gfp_flags = GFP_NOIO | __GFP_ZERO | __GFP_HIGHMEM | __GFP_THISNODE;
> page = alloc_pages_node(1, gfp_flags, 0);
EOF
cat >>/etc/systemd/system.conf <<EOF
CPUAffinity=numa
NUMAPolicy=bind
NUMAMask=0
EOF
cat >>/etc/memcached.conf <<EOF
-m 184320
-s /var/run/memcached/memcached.sock
-a 0766
-t 36
-B binary
EOF
cat fio.sh
modprobe brd rd_nr=1 rd_size=113246208
swapoff -a
mkfs.ext4 /dev/ram0
mount -t ext4 /dev/ram0 /mnt
mkdir /sys/fs/cgroup/user.slice/test
echo 38654705664 >/sys/fs/cgroup/user.slice/test/memory.max
echo $$ >/sys/fs/cgroup/user.slice/test/cgroup.procs
fio -name=mglru --numjobs=72 --directory=/mnt --size=1408m \
--buffered=1 --ioengine=io_uring --iodepth=128 \
--iodepth_batch_submit=32 --iodepth_batch_complete=32 \
--rw=randread --random_distribution=random --norandommap \
--time_based --ramp_time=10m --runtime=5m --group_reporting
cat memcached.sh
modprobe brd rd_nr=1 rd_size=113246208
swapoff -a
mkswap /dev/ram0
swapon /dev/ram0
memtier_benchmark -S /var/run/memcached/memcached.sock \
-P memcache_binary -n allkeys --key-minimum=1 \
--key-maximum=65000000 --key-pattern=P:P -c 1 -t 36 \
--ratio 1:0 --pipeline 8 -d 2000
memtier_benchmark -S /var/run/memcached/memcached.sock \
-P memcache_binary -n allkeys --key-minimum=1 \
--key-maximum=65000000 --key-pattern=R:R -c 1 -t 36 \
--ratio 0:1 --pipeline 8 --randomize --distinct-client-seed
Client benchmark results:
kswapd profiles:
5.19-rc1
40.33% page_vma_mapped_walk (overhead)
21.80% lzo1x_1_do_compress (real work)
7.53% do_raw_spin_lock
3.95% _raw_spin_unlock_irq
2.52% vma_interval_tree_iter_next
2.37% folio_referenced_one
2.28% vma_interval_tree_subtree_search
1.97% anon_vma_interval_tree_iter_first
1.60% ptep_clear_flush
1.06% __zram_bvec_write
patch1-6
39.03% lzo1x_1_do_compress (real work)
18.47% page_vma_mapped_walk (overhead)
6.74% _raw_spin_unlock_irq
3.97% do_raw_spin_lock
2.49% ptep_clear_flush
2.48% anon_vma_interval_tree_iter_first
1.92% folio_referenced_one
1.88% __zram_bvec_write
1.48% memmove
1.31% vma_interval_tree_iter_next
Configurations:
CPU: single Snapdragon 7c
Mem: total 4G
ChromeOS MemoryPressure [1]
[1] https://chromium.googlesource.com/chromiumos/platform/tast-tests/
Link: https://lkml.kernel.org/r/20220918080010.2920238-7-yuzhao@google.com
Signed-off-by: Yu Zhao <yuzhao@google.com>
Acked-by: Brian Geffon <bgeffon@google.com>
Acked-by: Jan Alexander Steffens (heftig) <heftig@archlinux.org>
Acked-by: Oleksandr Natalenko <oleksandr@natalenko.name>
Acked-by: Steven Barrett <steven@liquorix.net>
Acked-by: Suleiman Souhlal <suleiman@google.com>
Tested-by: Daniel Byrne <djbyrne@mtu.edu>
Tested-by: Donald Carr <d@chaos-reins.com>
Tested-by: Holger Hoffstätte <holger@applied-asynchrony.com>
Tested-by: Konstantin Kharlamov <Hi-Angel@yandex.ru>
Tested-by: Shuang Zhai <szhai2@cs.rochester.edu>
Tested-by: Sofia Trinh <sofia.trinh@edi.works>
Tested-by: Vaibhav Jain <vaibhav@linux.ibm.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com>
Cc: Barry Song <baohua@kernel.org>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Hillf Danton <hdanton@sina.com>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Michael Larabel <Michael@MichaelLarabel.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Mike Rapoport <rppt@kernel.org>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Qi Zheng <zhengqi.arch@bytedance.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Will Deacon <will@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Evictable pages are divided into multiple generations for each lruvec.
The youngest generation number is stored in lrugen->max_seq for both
anon and file types as they are aged on an equal footing. The oldest
generation numbers are stored in lrugen->min_seq[] separately for anon
and file types as clean file pages can be evicted regardless of swap
constraints. These three variables are monotonically increasing.
Generation numbers are truncated into order_base_2(MAX_NR_GENS+1) bits
in order to fit into the gen counter in folio->flags. Each truncated
generation number is an index to lrugen->lists[]. The sliding window
technique is used to track at least MIN_NR_GENS and at most
MAX_NR_GENS generations. The gen counter stores a value within [1,
MAX_NR_GENS] while a page is on one of lrugen->lists[]. Otherwise it
stores 0.
There are two conceptually independent procedures: "the aging", which
produces young generations, and "the eviction", which consumes old
generations. They form a closed-loop system, i.e., "the page reclaim".
Both procedures can be invoked from userspace for the purposes of working
set estimation and proactive reclaim. These techniques are commonly used
to optimize job scheduling (bin packing) in data centers [1][2].
To avoid confusion, the terms "hot" and "cold" will be applied to the
multi-gen LRU, as a new convention; the terms "active" and "inactive" will
be applied to the active/inactive LRU, as usual.
The protection of hot pages and the selection of cold pages are based
on page access channels and patterns. There are two access channels:
one through page tables and the other through file descriptors. The
protection of the former channel is by design stronger because:
1. The uncertainty in determining the access patterns of the former
channel is higher due to the approximation of the accessed bit.
2. The cost of evicting the former channel is higher due to the TLB
flushes required and the likelihood of encountering the dirty bit.
3. The penalty of underprotecting the former channel is higher because
applications usually do not prepare themselves for major page
faults like they do for blocked I/O. E.g., GUI applications
commonly use dedicated I/O threads to avoid blocking rendering
threads.
There are also two access patterns: one with temporal locality and the
other without. For the reasons listed above, the former channel is
assumed to follow the former pattern unless VM_SEQ_READ or VM_RAND_READ is
present; the latter channel is assumed to follow the latter pattern unless
outlying refaults have been observed [3][4].
The next patch will address the "outlying refaults". Three macros, i.e.,
LRU_REFS_WIDTH, LRU_REFS_PGOFF and LRU_REFS_MASK, used later are added in
this patch to make the entire patchset less diffy.
A page is added to the youngest generation on faulting. The aging needs
to check the accessed bit at least twice before handing this page over to
the eviction. The first check takes care of the accessed bit set on the
initial fault; the second check makes sure this page has not been used
since then. This protocol, AKA second chance, requires a minimum of two
generations, hence MIN_NR_GENS.
[1] https://dl.acm.org/doi/10.1145/3297858.3304053
[2] https://dl.acm.org/doi/10.1145/3503222.3507731
[3] https://lwn.net/Articles/495543/
[4] https://lwn.net/Articles/815342/
Link: https://lkml.kernel.org/r/20220918080010.2920238-6-yuzhao@google.com
Signed-off-by: Yu Zhao <yuzhao@google.com>
Acked-by: Brian Geffon <bgeffon@google.com>
Acked-by: Jan Alexander Steffens (heftig) <heftig@archlinux.org>
Acked-by: Oleksandr Natalenko <oleksandr@natalenko.name>
Acked-by: Steven Barrett <steven@liquorix.net>
Acked-by: Suleiman Souhlal <suleiman@google.com>
Tested-by: Daniel Byrne <djbyrne@mtu.edu>
Tested-by: Donald Carr <d@chaos-reins.com>
Tested-by: Holger Hoffstätte <holger@applied-asynchrony.com>
Tested-by: Konstantin Kharlamov <Hi-Angel@yandex.ru>
Tested-by: Shuang Zhai <szhai2@cs.rochester.edu>
Tested-by: Sofia Trinh <sofia.trinh@edi.works>
Tested-by: Vaibhav Jain <vaibhav@linux.ibm.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com>
Cc: Barry Song <baohua@kernel.org>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Hillf Danton <hdanton@sina.com>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Michael Larabel <Michael@MichaelLarabel.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Mike Rapoport <rppt@kernel.org>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Qi Zheng <zhengqi.arch@bytedance.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Will Deacon <will@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
